Yes, you're right that composing the best picture for an eye point could (and does) use splats from all over the scene.
But I think if you limit to splats that are (entirely, mostly, partially?) inside the 1m^3 block, you'll do pretty well. And you're absolutely right that reflective surfaces would probably be the first to suffer.
Well, it's worse than that. Because if you make a 1m^3 pond cube, and then I go putting trees around it, a naive rendering would still show YOUR reflections in the pond, rather than rendering from that pond's point of view, etc, like traditional rendering.
One of Gaussian Splats strengths, that it doesn't care... becomes a problem for me.
I was daydreaming about, picture if a corporation buys an entire state.
Say, Wyoming or West Virginia. Gemini guesses $180 billion to $250 billion.
With that investment, they'd get to control who lives in the state.
So, then the corporation gets to control who the Governor is. And the two Senators, and the seat in the Congress.
Forgive me if I'm wrong, but doesn't a single Senator have just a tremendous amount of power to block basically any legislation? With pocket vetoes, or silent filibusters?
Granted, actually buying a Senator is probably cheaper by a few orders of magnitude.
I blame the intentionally cheap receiving optics. Researchers have reconstructed a TV signal from a hotel room with closed blinds, by very accurately measuring the ambient light intensity of the room (as shown on the blinds.)
Knowing that, I doubt that someone with even moderate funding would have difficulty receiving a signal from any of the transmitters you mention.
But, in all honesty, if you put a physical cap on the transmitter and receiver, maybe I'm wrong.
But in the other hand, none of the devices I currently own have one of your transmitters, and they all have screens and cameras, so...
Thanks for the dialogue and for sharing your experiences.
I specifically want this device to have no input or output hardware that could be used without my knowledge. IrDA could absolutely be used without my knowledge.
Me, in my life, I have a PC that's connected to the internet. I have a phone that's connected to the internet.
I want another device, which I imagine to be a Pi or Esp32 or something with a camera and a touchscreen display, and capacitive charging. After I program it and give it the public/private keypair and the OTP, I imagine physically breaking off a USB port, or sealing one with some hardening resin.
I don't want an entire airgapped computer. Maybe you do, that's fine. For me, I'd love it to be a credit-card sized doodad.
I've wanted to use this for an air-gapped communication device.
I have a device with a camera and a touch-screen that only uses capacitive charging. I type a message. Bytes are encrypted. I hit send. QR codes flash on my screen. I use my PC or my normal phone to receive the encrypted bytes, and transmit them to you. You have the same device. You have your PC or phone flash encrypted QR codes. You use your device to receive, and then decrypt.
I've daydreamed about also buying several different hardware random noise generators. XOR all of their bits together. Save a huge one time pad to each of our devices. And then also use public key crypto on top of it.
I'm not really sure why I want this. But, it's my answer for how to reduce attack surface as much as possible, and have truly secret messages.
I'll admit that, at first, I thought this said Michael Keaton. And oddly enough, Michael Keaton's real name is Michael Douglas. So, there's a whole spiral of madness that my brain went off into.
If I'm trying to tour one of the beautiful servers that people have made, yes, it absolutely would work for my use case.
I'm sorry my idea doesn't appeal to you.
Acting like it has no value for anyone means you don't understand my use case, or you don't care about me. It may entirely be my fault that you don't understand my use case, but I think you're being obstinate.
I've been toying with using Jart's Blink, the "tiniest x86-64-linux emulator". If it were hardened against attacks by a smart community, it seems really attractive...
Then I've been toying with using WASM, with a strictly limited API surface... (API stands for what, "Attack, Please, Infiltrators"?)
I'd really love to make a service which can run "untrusted" code... I want this to exist. I want it to feel like the Sandstorm.io marketplace, and to use Tailscale or something like it... and then to make "Facebook but only for friends or friends-of-friends." No public figures. No random people or ads in my feeds. Just friends. And then other similar apps - mail, discord, etc. Don't let it make outbound HTTP connections. Don't let it access local filesystem, other than assets that were bundled with it. Give it a SQLite API... And let it handle inbound HTTP, REST, WebSocket requests... And use gRPC over Tailscale to only talk directly to approved contacts...
Thoughts? How to run untrusted code safely?
I tried using Firecracker and gVisor... But I'm dissatisfied with them. I was trying to make it easy for someone to download and run my service... But Firecracker and gVisor don't at all feel like "libraries that install cleanly." They feel more like, if you're enterprise scale, you can hire someone to make sure Firecracker and gVisor work the way you want... Maybe I'm just not trying hard enough... Maybe using WSL2 on Windows 11 is just asking for trouble. I tried several different ways to run Linux on Windows, and none of them seemed very happy, when I tried to add Firecracker / gVisor.
Yeah, when you describe that, I picture Wave Function Collapse to generate a map schematic... And then a text prompt, and some style photos the designers want it to match.
I want some math Phds to sit down with the Corridor Crew guys and figure out how to make Gaussian Splat reconstruction way better, the way they came up with CorridorKey.
I know feelings about AI are mixed. But when AI can dream up gaussian splats in real time, from a prompt, and do refinement as you get closer to things... That's going to be pretty bonkers.